obtained, normal DMKMCH processing continues. If the lock cannot be
obtained, DMKftCH stacks a CPEXBLOK with CPftCBLK set and exits to DMKDSPRU. This CPEXBLOK causes processing to resume at DMKftCHSE with the
global system lock held. Any machine checks that occur before the CPEXBLOK processing has completed are considered recursive machine checks and handled appropriately. If the control program was in control
and the system is not to be terminated, the machine check handler saves
status in the CPEXBLOK, set CPMCHLK and reloads MCOPSW. CPftCHLK is set
to prevent the dispatcher from starting any new work on this processor
until the machine check processing has completed. DMKMCH passes control to DMKftCTPT if the system is running in
attached processor mode and a decision has been made to terminate the system. In general, if a virtual machine was running when the machine chsck occurred, only that virtual machine is terminated. DMKMCTPT determines if the system can continue and if the processor
can continue. For the attached processor, if the machine check was not
a clock error and the control program was not in control, the virtual machine running at the time of the error is terminated. If the machine check was a clock error on the main processor or the control program was in control on either processor, the other processor is signalled to stop
and store status and a wait state PSW is loaded on the failing
processor. An attempt is made to issue message 610W to the operator
before the main processor is stopped. If the machine check was a clock
error on the attached processor and the control program was not in
control, the main processor is signalled via an external call to
initiate automatic processor recovery with an indicator to continue
processing.
The malfunction alert interrupt handler (DftK!CTftA) receives control from the external second level interrupt handler. If the malfunction
alert came from the main processor, a 001 wait state PSW is loaded. If
the malfunction alert came from the attached processorand a virtual machine was in control, an indication is set to terminate the virtual
user and CPAPBPND is set for processor recovery. If the attached
processor was in supervisor state, message 610W is sent to the operator
and a 013 wait state PSi is loaded. If the attached processor was in a wait state, CPAPRPND is set for processor recovery.
The automatic processor recovery routine (DftKMCTPB) receives control
from the external SLIH or the dispatcher. If the system is to continue
processing, the vary processor offline routne (DftKCPUUP) is called. DMKCPUUP examines the chain of virtual machines for attached processor
affinity and shared segment pointers. Any shared segment pointers for
the attached processor are switched to point to the main processor
shared segments. All the system control blocks and save areas necessary
to run in attached processor mode are also freed. The time froa the
first timer request queue element is placed into the clock comparator
for the main processor. While preserving the maintained fields in the absolute zero area, the main processor's prefix storage area is copied to the absolute zero area
and prefixing is stopped. The attached processor operational flag is turned off in the absolute zero area, and the prefix storage areas for
the main and attached processors are freed. The pages and DISD slots
held by the attached processor for shared segments are freed by D!KPGT and DMKPTR. A message (194I) is issued, and return is made to DftK!CTPR. For any virtual machines with affinity to the attached processor, DMKMCTPR resets the affinity for each, issues message 6211, and puts the
machine in console function mode (if the virtual machine is not
disconnected). If a virtual machine is to be terminated, the virtual machine is reset, messages 616I and 6191 Normal return canses the system to continue processing in uniprocessor mode. 1-176 IBM VM/370 System Logic and Problem Determination--Volume 1

The action that. the machine check handler takes for a given situation
is determined by the error itself,' the operating environ.ent of VM/370, and whether the system was performing a CP function or a virtual machine function --or the system was not performing at all (a loaded wait state
condition when the error occurred). Figure 24 clarifies the action the
system takes for the gi ven si tua tions. I VM/370 Processing IVirtual Machine processing I I I I I Error Condition I uniprocessor attached processor I uniprocessor attached processor I I I I I Main Attached Main I attached I I I I Invalid machine I 1 1 1 1 1 1 I check interrupt I code Invalid PSi 1 1 1 1 3 3
data
Register, 1 1 1 3 3 Program Ilask instruction
address invalid
System damages 1 1 1 1 3 3
TOD or CPU 1 1 1 1 1 3,4
Clock Errors
Multibit 1 1 1 3,2 3,2 3,2 (solid) Storage error
Multibit 1 1 1 3,2 3,2 3,2
(intermi ttent)
storage error Storage Protect 1 1 1 3 3 3 Key (solid)
failure Storage Protectl 2 2 2 2 2 2
(intermittent) I failure I I Malfunction I 5 1 1 5 1 3,4
alert I Channel 1 1 1 1 1 1
inoperative Legeng: I 1 = load wait state PSi I 2 = refresh for retry operation I 3 = terminate the virtual .achine I 4 = automatic processor recovery I 5 = Not applicable
Figure 24. Condition/Action Table for Uncorrectable Errors CP Introduction 1-177